Goal: Load Twitter API into a data management system and write queries to retrieve data.

  • Dataset of Elon Musk’s most recent Tweets during 2015-2022, stored in csv format, where each tweet is in its own separate row object.
user_id screen_name followers_count friends_count account_created_at verified
x44196397 elonmusk 95589999 114 2009-06-02 TRUE

Introduction

Apache Spark is an open-source engine for large-scale parallel data processing known for its speed, ease of use, and cutting-edge analytics. It provides high-level APIs in general-purpose programming languages such as Scala, Python, and R, as well as an optimization engine supporting standard data analysis methods.

Azure Databricks is an analytics platform based on Microsoft Azure cloud services, providing the latest versions of Apache Spark and allowing consistent integration with open source libraries. Built with Spark capabilities, Databricks provides a cloud platform with an interactive workspace and fully managed Spark clusters, allowing users to work in a single, easy-to-use environment, create and configure clusters in seconds, and quickly execute Spark code.

Data Description

Twitter (Elon Musk 2015-2022): Dataset of Elon Musk’s most recent Tweets during 2015-2022, stored in RDS format, where each tweet is in its own separate row object. All Tweets are collected, parsed, and plotted using rtweet in R. In total, there are more than thousands of tweets in this dataset, including retweets and replies. All objects are to go into a single database.

Data Set Variables
1 status_id 14 hashtags 27 quoted_followers_count 40 retweet_location
2 created_at 15 symbols 28 quoted_location 41 retweet_description
3 user_id 16 media_expanded_url 29 quoted_description 42 retweet_verified
4 screen_name 17 media_type 30 quoted_verified 43 name
5 text 18 mentions_screen_name 31 retweet_status_id 44 location
6 source 19 quoted_status_id 32 retweet_text 45 description
7 reply_to_screen_name 20 quoted_text 33 retweet_created_at 46 followers_count
8 is_quote 21 quoted_created_at 34 retweet_source 47 friends_count
9 is_retweet 22 quoted_source 35 retweet_favorite_count 48 account_created_at
10 favorite_count 23 quoted_favorite_count 36 retweet_retweet_count 49 verified
11 retweet_count 24 quoted_retweet_count 37 retweet_user_id
12 quote_count 25 quoted_user_id 38 retweet_screen_name
13 reply_count 26 quoted_screen_name 39 retweet_followers_count

Twitter API

Once you have your twitter app setup and loaded the rtweet package in R, you can run the following command to generate a token for the Twitter data. The first thing that you need to setup in your code is your authentication and then create a token that authenticates access to tweets.

library(rtweet) # load rtweet package

twitter_token <- create_token(app = "mytwitterapp",
  consumer_key = "api_key", consumer_secret = "api_secret_key",
  access_token = "access_token", access_secret = "access_token_secret")

use_oauth_token(twitter_token) # authenticate via web browser

Now we search Twitter’s full archive API. We run the search_fullarchive command in order to be able to access historical tweets from a user. The example below captures all of Elon Musk’s tweets from January 01, 2010 to May 28, 2022.

df <- search_fullarchive(q = "from:elonmusk", n = 10000, env_name = enviroment_name, fromDate = "201001010000", toDate = "202205280000")
Data Set Preview:
created_at screen_name text favorite_count retweet_count quote_count reply_count is_quote is_retweet
2022-05-27 22:52:53 elonmusk Starship 24 rolls out to the pad at Starbase https://t.co/PGh6FY6x8w 0 0 0 0 FALSE TRUE
2022-05-27 22:16:31 elonmusk (westcoastbill?) It was 96% in favor of me, until I commented. Ironically, there might be more left-wing bots following me than right-wing haha. 23603 1177 83 1466 FALSE FALSE
2022-05-27 22:03:30 elonmusk (IheartTesla?) (Tesla?) V true 1869 92 7 152 FALSE FALSE
2022-05-27 20:59:20 elonmusk (EvaFoxU?) Yes 3014 142 9 236 FALSE FALSE

Questions of Interest

  1. (path finding) Display the thread (replies) of tweets (the tweet, time, id, in reply to id, user name with their screen name) posted by Elon Musk with screen_name in the order in which they were posted.
  2. (location) From which location have the tweets been most actively posted (most number of tweets)?
  3. (hashtags) Which hashtags does Musk use the most, and how many tweets are associated with these hashtags?
  4. (topics) What word does Musk mention the most in his tweets? What company products does Musk mention the most in his tweets? Products include Falcon 9, Starlink Satellites, Model 3 cars, etc.
  5. (trending) Are there any trends of what Musk tweets about the company?
  6. (nature of engagement) What is the percentage of different types of tweets (simple tweet, reply, retweet, quoted tweet) to their overall number of tweets?

Storing Data in Databricks

Building a Databricks workspace using an Apache Spark cluster.

First, we create a Databricks workspace from the Azure portal and then launch the workspace, which redirects us to the interactive Databricks portal. We create a Spark cluster from the Databricks interactive workspace and configure a notebook on the cluster. In the notebook, we can use either PySpark or SparkR to read data from a dataset into a Spark DataFrame. Using the Spark DataFrame, we can run a Spark SQL job to query the data.

The Azure Databricks system stores this data file in FileStore, located in the FileStore/ folder, which we then use the following PySpark or SparkR command to read in the data as a dataframe.

Python Code: 

df = spark.read.csv(path = 'dbfs:/FileStore/dfclean.csv',
                    header = "true", multiLine = "true")

# register table so it is accessible via SQL Context
df.createOrReplaceTempView('twitterTemp')

R Code: 

require(SparkR)
df <- read.df("dbfs:/FileStore/dfclean.csv", source = "csv", 
              header="true", inferSchema = "true")
createOrReplaceTempView(df, "twitterTemp") # register table

FileStore is a specific folder in DBFS where users can import and store data files from their local computer into a DBFS web browser. The Spark API then reads the imported data into an Apache Spark DataFrame as shown above.

Querying Twitter Data

Users can perform relational procedures on DataFrames using a domain-specific language (DSL) similar to R dataframes and Python Pandas. DataFrames support standard relational operators, including projection (select), filter (where), join, and aggregations (group by).

Query 1. Mentions

  • Display the thread (replies) of tweets (the tweet, time, id, in mention to id, user name with their screen name) posted by Elon Musk with screen_name in the order in which they were posted.

Python Code: 

mentionDF = spark.sql("SELECT created_at, text, \
    mentions_user_id, mentions_screen_name \
    FROM twitterTemp")
mentionDF.filter(mentionDF.mentions_user_id != "NA")
mentionDF.createOrReplaceTempView('mentionDF')

R Code: 

require(SparkR)
mentionDF <- select(df, "created_at", "text",
    "mentions_user_id", "mentions_screen_name")
mentionDF <- filter(mentionDF, mentionDF$mentions_user_id != "NA")
createOrReplaceTempView(mentionDF, "mentionDF")

 

created_at text mentions_user_id mentions_screen_name
2011-12-21 11:13:59 Yeah, this really is me, as my Mom (mayemusk?) will attest. Not sure I can handle just doing 140 char missives. Will put longer thoughts on G+ x14171401 mayemusk
2011-12-27 19:54:15 Vending Machine Attendant Admits B3 Selection Has Changed A Lot Over The Years http://t.co/nccSGzCQ #OnionInnovation x14075928 TheOnion
2011-12-28 22:29:48 Any Idiot Could Have Come Up With The Car http://t.co/e9cLgfEg #OnionInnovation x14075928 TheOnion
2011-12-28 22:27:08 (TheOnion?) So true :) x14075928 TheOnion

SQL Query: 

SELECT mentions_screen_name,
  COUNT(*) AS n
FROM mentionDF
WHERE mentions_screen_name != 'NA'
GROUP BY mentions_screen_name
SORT BY n DESC;

Query 2. Types

  • What is the percentage of different types of tweets (simple tweet, reply, retweet, quoted tweet) to their overall number of tweets?

When creating a managed table, Spark will manage both the table data and the metadata. To obtain the percentage of different types of tweets, we begin by creating a managed table for each tweet type (simple, reply, retweet, quote). The following R command returns Elon Musk’s different tweet types and overall frequency.

R Code: 

retweets <- nrow(df[df$is_retweet == "TRUE", ])
quotes <- nrow(df[df$is_quote == "TRUE", ])
normal <- nrow(df[df$is_quote == "FALSE" & df$is_retweet == "FALSE", ])
replies <- nrow(df[df$reply_to_status_id != "NA", ])

data <- data.frame(type = c("retweet", "quote", "regular", "reply"),
  n = c(retweets, quotes, normal, replies))

typeDF <- createDataFrame(data)
createOrReplaceTempView(typeDF, "typeDF")

So the different tweet types are retweet, quoted, reply_to, and normal.

Now, let’s create a data table to display the information for each of Elon Musk’s retweets and query the results to obtain the number of times Musk retweeted each user. With these results, we can visualize the most frequent twitter users Musk retweets as shown in the above graphic report.

CREATE TABLE retweets (
  retweet_status_id CHAR, 
  retweet_text TEXT, 
  retweet_created_at DATE,
  retweet_source TEXT,
  retweet_favorite_count INT,
  retweet_retweet_count INT,
  retweet_user_id CHAR,
  retweet_screen_name TEXT,
  retweet_name TEXT,
  retweet_followers_count INT,
  retweet_friends_count INT,
  retweet_statuses_count INT,
  retweet_location TEXT,
  retweet_description TEXT,
  retweet_verified BOOLEAN
);

# Load data into table
LOAD DATA INPATH '/user/dfclean' INTO TABLE retweets;

Retweets Table:

retweets
created_at text screen_name favorite_count retweet_count followers_count location description verified
2022-04-27 23:41:36 Docking confirmed! https://t.co/RyPZBAv5Lo SpaceX 128109 6890 22548863 Hawthorne, CA SpaceX designs, manufactures and launches the world’s most advanced rockets and spacecraft TRUE
2020-05-14 23:22:56 Breakthrough, Part Deux (LVCVA?) https://t.co/JQ7sjPXSfZ boringcompany 7278 649 849005 Austin/Las Vegas Beat the snail TRUE
2016-12-05 06:33:41 We’re releasing Universe, a platform for measuring and training AI agents: https://t.co/bx7OjMDaJK OpenAI 3088 1759 748401 NA OpenAI’s mission is to ensure that artificial general intelligence benefits all of humanity. We’re hiring: http://openai.com/jobs TRUE
2016-03-31 20:10:58 Longest line EVER #Model3 https://t.co/EafMhK3BDC Tesla 5487 3059 15719350 NA electric cars, giant batteries and solar TRUE

Query 4. Hashtags

  • Which hashtags does Musk use the most, and how many tweets are associated with these hashtags?

Here we want to extract hashtags from the content of the Tweets data. The following command unpacks the hashtags column into an array of strings, followed by counting how many unique hashtags used by Elon Musk.

SQL Query: 

SELECT hashtags, COUNT(*) AS hashtagcount
FROM twitterTemp
WHERE hashtags != 'NA'
AND hashtags != '0'
GROUP BY hashtags
SORT BY hashtagcount DESC;

Query 5. Topic Words

  • What word does Musk mention the most in his tweets? What company products does Musk mention the most in his tweets?

Python Code: 

import pyspark.sql.functions as f
textDF = df.select('text', 'status_id', 'created_at')

# Count and group word frequencies on text, when split by space comma
textDF.withColumn('word', f.explode(f.split(f.col('text'), ' '))) \
  .groupBy('word') \
  .count() \
  .sort('count', ascending=False)

R Code: 

reg <- c("&amp;|&lt;|&gt;", "[^\x01-\x7F]",
         "\\s?(f|ht)(tp)(s?)(://)([^\\.]*)[\\.|/](\\S*)")

df %>% select('text', 'status_id') %>%
  mutate(text = str_remove_all(text, reg)) %>%
  unnest_tokens(word, text, token = "tweets") %>%
  filter(!word %in% stop_words$word,
         !word %in% str_remove_all(stop_words$word, "'"),
         str_detect(word, "[a-z]"),
         !str_detect(word, "^#@\\S+")) %>%
  count(word, sort = TRUE)

Sentiment Analysis

Here we use the syuzhet R package to iterate over a vector of strings consisting of the text from all of Elon Musk’s tweets in our dataset. To obtain the vector of tweet text, the plain_tweets() function from the rtweet package is used to clean up the tweets character vector to cleaned up, plain text. We then pass this vector to the get_sentiment() function, which consequently returns the sentiment values based on the custom sentiment dictionary developed from a collection of human coded sentences.

R Code: 

round_time <- function(x, secs)
  as.POSIXct(hms::round_hms(x, secs))
sent_scores <- function(x)
  syuzhet::get_sentiment(plain_tweets(x)) - .5

df.sentiment <- gfg_data %>%
  dplyr::mutate(days = round_time(created_at, 60 * 60 * 24),
                sentiment = sent_scores(text)) %>%
  dplyr::group_by(days) %>%
  dplyr::summarise(sentiment = sum(sentiment, na.rm = TRUE))

ext Mining: Sentiment Analysis Once we have cleaned up our text and performed some basic word frequency analysis, the next step is to understand the opinion or emotion in the text. This is considered sentiment analysis and this tutorial will walk you through a simple approach to perform sentiment analysis.

To perform sentiment analysis we need to have our data in a tidy format. The following converts all seven Harry Potter novels into a tibble that has each word by chapter by book. See the tidy text tutorial for more details.

Using the same package, we now use the get_nrc_sentiment() function to get the emotions and valences from the NRC sentiment dictionary for each tweet. First, we must clean the text column from our dataset

This syuzhet function calculates the presence of emotions and their valence for each tweet, where the text is cleaned so that each row holds a sentence from our dataset.

The columns include one for each emotion type as well as a positive or negative valence. The ten columns are as follows: “anger”, “anticipation”, “disgust”, “fear”, “joy”, “sadness”, “surprise”, “trust”, “negative”, “positive.”

R Code: 

cleanTweet  = as.vector(df$text)
cleanTweet = gsub("rt|RT", "", cleanTweet)
cleanTweet = gsub("http\\w+", "", cleanTweet)
cleanTweet = gsub("<.*?>", "", cleanTweet)
cleanTweet = gsub("@\\w+", "", cleanTweet)
cleanTweet = gsub("[[:punct:]]", "", cleanTweet)
cleanTweet  = gsub("\r?\n|\r", " ", cleanTweet)
cleanTweet = gsub("[[:digit:]]", "", cleanTweet)
cleanTweet = gsub("[ |\t]{2,}", "", cleanTweet)
cleanTweet = gsub("^ ", "", cleanTweet)
cleanTweet = gsub(" $", "", cleanTweet)

textSentiment <- get_nrc_sentiment(cleanTweet)
nrc_sentiment <- cbind(df, textSentiment) %>% 
  dplyr::select(created_at, anger, anticipation, disgust, fear, 
                joy, sadness, surprise, trust, negative, positive)

References

Brown, Leif, Jason Howell, and Mary McCready. 2022. “Quickstart - Run a Spark Job on Azure Databricks Workspace Using Azure Portal.” Microsoft Technical Documentation. https://docs.microsoft.com/en-us/azure/databricks/scenarios/quickstart-create-databricks-workspace-portal.
Kearney, Michael W. 2019. “Rtweet-Workshop.” Data Science and Analytics Presentation. University of Missouri School of Journalism: Informatics Institute. https://rtweet-workshop.mikewk.com/.
Kearney, Michael W., Francois Briatte, Andrew Heiss. 2019. “Rtweet: Collecting and Analyzing Twitter Data.” Journal of Open Source Software 4 (42): 1829. https://doi.org/10.21105/joss.01829.